Class SlidingWindow
-
template<typename NT>
class SlidingWindow Sliding window for tracking convergence based on relative error Maintains a fixed-size window of recent values to detect when optimization has converged by comparing the oldest and newest entries.
- Template Parameters:
NT – Numeric type for stored values
Public Functions
-
inline SlidingWindow(int windowSize)
Construct a sliding window with specified size
- Parameters:
windowSize – [in] Maximum number of values to store
-
inline void push(NT value)
Add a new value to the window Adds value to the front of the window. If window is full, removes the oldest value from the back.
- Parameters:
value – [in] The new value to add
-
inline NT getRelativeError() const
Calculate relative error between newest and oldest values Computes |oldest - newest| / |oldest| to measure convergence. Returns 1.0 if window not full or to avoid division by zero.
- Returns:
Relative error in range [0, infinity), or 1.0 if not converged
-
inline bool isFull() const
Check if window is full
- Returns:
True if window has collected windowSize entries, false otherwise
-
inline void clear()
Clear all values from the window Resets the window to empty state.